Emitting marker names instead of hex statement addresses for SDT probes?

[Craig Ringer <craig@2ndquadrant.com>]

Hi all

I'm looking for advice on how feasible it might be to have systemtap report
the SDT marker name in the probe-point name, instead of a
.statement(0xdeadbeef) string.

Because probes like process("foo").mark("bar") are converted to
process("foo").statement(0xdeadbeef), the monitor mode display is nearly
useless for SDTs. You just get a wall of undifferentiated hex statement
addresses. It also makes various errors and diagnostics much more difficult
to act on, and makes the output of pp() nearly useless for SDT markers.

Any advice on where I might start looking in the code if I wanted to change
this?

Also, ideally I'd also like to have $$name do something sensible for all
probe point types, so I don't need to

    defined($$name) ? $$name : ppfunc()

in macros and functions that can be invoked by both marker based and
DWARF-based probes.

Any thoughts on the most sensible value to assign to $$name for other probe
types? It could resolve to pp(), but that can be very long and verbose.
ppfunc() would be logical for function based probes but for others it's
less obvious.




-- 
 Craig Ringer                   http://www.2ndQuadrant.com/
 2ndQuadrant - PostgreSQL Solutions for the Enterprise